Every year at the Santa Fe Institute we explore one or more themes touching on SFI science. In 2018 we are focusing on the complexity of intelligence, both natural and artificial.
SFI plans theme years through a series of small discussion meetings—which we call sherpa meetings—in each preceding year that establish a range of topics for working groups, pinpoint key challenges facing a field, and most importantly, identify individuals with novel, intriguing, or radical ideas.
Sherpa meetings are small (around 10-15 participants each), informal (prepared talks are passionately discouraged), and lively (strong opinions are appreciated).
Artificial and Natural Intelligence
A defining feature of complex systems is their ability to encode, store, process and employ functional information. This feature encompasses the elementary gradient sensing capabilities of single cells, through to the large-scale perceptual and decision-making abilities of large populations of neurons.
Intelligent systems employ both intrinsic and extrinsic mechanisms of storage and processing to efficiently solve adaptive challenges. These solutions include volatile chemical signals supporting ant foraging, bee hives coordinating collective navigation, through to a vast variety of human inventions.
The ecologically grounded limitations of organically evolved natural intelligences has lead to pressures in favor of the development of culturally evolved artificial intelligences. The most obvious examples of these limitations are the constraints of working memory and the shortcomings of rudimentary mental arithmetic.
In technological settings many features associated with intelligent behavior are outsourced to artifacts that serve as a community resource. These include language, mathematics, calculators, modern digital computers, and a growing library of inferential software.
This meeting seeks to explore hybrid natural and artificial intelligence, in which representations, storage, and inference are shared across organically and culturally evolved intelligent systems (from spider webs to the WWW). This investigation raises a number of critical questions
1. What are species unique and species spanning forms of intelligence? And how can these be measured and compared?
2. How should we think about the differences and commonalities among natural perceptual, motor and “analytical” intelligence?
3. Is there a general intelligence that supports all of these functions or is intelligence modular in a fundamental sense?
4. Which features of organic intelligence and its modules lend themselves to artifactual and artificial amplification and replacement?
5. What novel forms of A.I can be derived from a careful consideration of the full diversity of N.Is?
This meeting has been generously supported by Bill Miller through the Miller Omega Fund.